Abstract
Limiting the global warming impact requires a drastic reduction of greenhouse gas emissions (GHG) to reach a zero-carbon growth by 2050. Against this backdrop, turbomachinery designers are asked to develop more compact, efficient, and reliable machines. The research focuses on developing machines able to reduce energy losses and overall footprint, thus improving power plants efficiency and operability. To this end, expander-compressors (EC) seem to play a key role in the energy transition and appear a promising technology for high-density plants, especially together with hydrogen (H2) or supercritical carbon dioxide (sCO2) as working fluid. However, when operating conditions changes, a reduction in machine efficiency may be expected, with a consequent degradation of the overall system performance. In order to optimize the operation of the power plant, it is crucial to understand how the variation of the expander-compressor behavior affects the working conditions of the other components of the plant. In this perspective, a fast and reliable way to estimate the EC performance at different operating conditions is necessary. Although the scientific literature shows several studies on mean-line approaches for expanders and centrifugal compressors, the joint use of 1D models with artificial neural networks (ANN) for ECs combined machine is still overlooked. To fill this gap, the prediction of artificial intelligence (AI)-based performance maps is proposed in this work. The tool has been tested in a case study based on a real EC. Results showed a rapid evaluation of EC efficiency while varying the operating conditions.
